User:Gmaxwell/An alternative model for license compatibility
For reasons outside of the scope of this document it is highly desirable to enable compatibility between the two most widely used copyleft content licenses: CC-BY-SA, administered by the Creative Commons, and the GFDL, administered by the Free Software Foundation (FSF).
When we speak of compatibility, we are primarily interested in the ability to create derivative works combining source material from both licenses.
Progress to date
A commonly suggested solution involves changing the GFDL to include a clause such as "You may, alternatively, distribute verbatim copies or derivatives of this work under CC-BY-SA". Thus anyone seeking compatibility could first 'convert' the GFDL content to CC-BY-SA and then combine freely.
If the CC-BY-SA incorporated equivalent terms, content could move freely between the licenses.
It has been argued by some, such as Larry Lessig, that such functionality would provide beneficial immunity to license flaws. However, the protection actually provided may not actually be useful.
Flaws in a copyleft license can take two primary forms: that of being too restrictive, and that of being too permissive. Although many flaws can be considered a mixture of these two factors, the factors provide a useful method for breaking down and understanding license flaws.
Flaws of restrictiveness are flaws which prohibit beneficial uses of the free content disproportionate to the restrictions' benefit for the preservation and promotion of free content. It is widely argued that the requirement in GFDL 1.2 to include the entire license (many kilobytes of text) with GFDL licensed content, even the smallest possible copyrighted excerpt, is a flaw of excess restrictiveness.
Flaws of permissiveness are flaws which allow behavior counterproductive to the preservation and promotion of free content. For example, the CC-BY-SA does not require content users to include the actual license terms electronically, in favor of a human-opaque URL to the Creative Commons website, even when copying gigabytes of covered works. This is seen by some as a failure of permissiveness, because rights that the public is unaware of are largely meaningless rights.
Flaws of restrictiveness can always be fixed retroactively because of the upgrade clauses in these licenses: a new version of the license can be released which relaxes the restriction. Furthermore, since these licenses are already fairly permissive it is likely that new flaws of restrictiveness will arise only at a fairly slow pace as new modalities of communication emerge with unexpected implications.
Flaws of permissiveness can not be fixed retroactively. Someone wishing to engage in the unwanted behavior can continue to use the old license. This is an unfortunate but necessary side effect of the irrevocability of these licenses which is fundamental to their free nature. A license update can only solve the permissiveness problem for new whole works or derivatives. Because copyleft licenses are intended to be fundamentally permissive, flaws of permissiveness are infrequent. However because permissiveness flaws are often discovered by motivated intelligent attackers (rather than the accidental invention of disruptive new distribution models), it would not be surprising to find them discovered on a more frequent basis.
The aforementioned license mobility solution maximizes security against flaws of restrictiveness, but minimizes security against flaws of permissiveness. Since the implications of a permissiveness flaw are more severe, and because restrictiveness flaws can be so easily corrected it is apparent that this solution is optimizing for the wrong case.
Compatibility is created, but only at an increased risk of excessive permission.
An alternative model
One possible alternative model for license compatibility would be to adjust the licenses to allow distribution under the union of the licenses. This is somewhat similar in spirit and operation to the approach taken in the GPLv3 drafts for license compatibility, in that we would allow optional added restrictions which could not later be removed in downstream versions.
This could be achieved first by producing new versions of the CC-BY-SA and the GFDL which eliminate mutually exclusive terms, and by addressing any restrictiveness flaws seen in both licenses, compromising where necessary.
An example compromise might have the SFDL terms require the complete text of the license only need to be included only if the total aggregate GFDL-covered work (including FDL/CC-BY-SA mixtures) crosses a size threshold which is far above the size of the license itself. (Such a change already exists in the SFDL, and since license-tagalong is an acceptable restriction under the Free Content definition, it is likely the CC does not find such restrictions too offensive.)
A possible lineage of licenses for this scenario is provided by the illustration on the right.
The new GFDL is called the SFDL, or "Simplified Free Documentation License", reflecting the larger changes which may be required in this license and that fact that the FSF has already drafted a SFDL which goes a long way towards making the needed changes.
In this example, both the SFDL-3.0 and CC-BY-SA-3.0 would contain terms specifically allowing content under the other license to be mixed, so long as the restrictions of both licenses were applied to the result. These additional restrictions would be permitted in this context.
New versions of each license could continue to be released, but only when both licenses have updated can combined content move forward to another version. Non-combined content is unaffected by this compatibility.
This model is maximally weak against flaws of restriction: a bad restriction in either license leaves the combined works impacted. However, this model is maximally strong against flaws of permissiveness, as only when both licenses are excessively permissive are the combined works impacted. Because flaws of restrictiveness are able to be totally solved after the fact, it is likely that most would find this failure mode preferable over the long term.
Some have argued that the upgrade clauses in both the CC-BY-SA and the GFDL are potentially a flaw of permissiveness. The upgrade clauses allow the body controlling the license to make changes which the copyright holders of the license may find undesirable. However, most would also agree that the upgrade clause is an important protection against the volatile nature of the real world.
The alternative model proposed here may provide a powerful protection against flawed actions by the bodies controlling these licenses. In this model each work in the pool of existent copyleft content exists in one of three states: SFDL, SFDL+CC, or CC-BY-SA. If these largely independent licensing bodies are unable to agree, the combined content remains under the old licenses until an agreement is reached. This provides a natural balance controlling the actions of both parties.
For example, if in CC-BY-SA the Creative Commons introduced a new term forbidding derivative works from being used to ridicule an upstream copyright holder, the Free Software Foundation would likely believe that this would be an unacceptable limitation on the future use of a free content work and would decline to create a matching compatible version of the GFDL.
Content purely under the CC license and purely under the GFDL could go their separate ways, but content which combined material under these licenses would be stuck at an older revision level until the licensing parties came to an agreement. A mutual interest in avoiding confusion would encourage both parties to only use equal version numbers for compatible versions.
Although this process will no doubt slow the introduction of new versions, the need for new versions is seldom urgent, and update-causing events will likely be spread further apart in time as the licenses and our understanding of the needs of free content mature.